Supervised Learning - Prediction Week 3 Challenge

2Ans:

Mean Absolute Error is measure the average magnitude of the error in a set of predictions,without considering their direction. It is the average over the test sample of the absolute difference between prediction and actual observation where all individual difference have equal weight.

Root Mean Square Error is an quadratic scoring rule that also measure the average magnitude of the error. It is the square root of average squared difference between prediction and actual obseration. Root Mean Square gives a relatively high weight to large error. This means the root mean square should be more useful when large error are particularly undesirable.

3Ans:

Regression involves prediction a specific value that is continues in nature. Estimating the price of a house or predicting a specific stock prices are examples of regression becasuse one works towards building one would predict a real valued quantity.

Mean Absolute Error : It is one of the most simple yet robust loss function used for regression model. Regression problems may have variables that are not strictly guassion in nature due to the presence of outliners.

Mean Absolute Error would be an ideal option in such cases becauses it does not take into account the direction of the outliners.

1. ii) Mean Square Error : It is the average of the squared difference between the actual and prediction values.

iii) Mean Bias Error : It is the used to calulate the average bias in the model. Bias in the nutshell is overestimating or underestimating a parameter.

Mean Bias Error takes the actual difference between the target and the predicted value and not absolute difference one has to be caution as the postive and the negative errors could cancel each otherout which is why it is one of the lesser used loss function.

iv)Mean Squared Logarithms Error : Calculating the mean squared logarithms error is the same as mean squared Error except the natural logarithm of predicted values is used rather than the actual values.

v)Huber Loss: It combines the robustness of L1 and with the stability of L2 essentially the best of L1 and L2 losses for huge errors it is the linear and for small errors it is quadratic in nature.

4Ans :

In Machining Learning we deal with two types of parameters

1. i) Machine Learnable Parameter
2. ii) hypyer Parameter
3. i) Machine Learable parameter are the one which the algorthim learn estimate on their own during the training for a given dataset. β0,β1,β2, are the machine learnable parametres.
4. ii) Hyper paramater are the one which the machine learnable engineers or the data scientists will assign specific values to the controls the way of alogarithms learns and also to the perforamance of the model. Learning rate generally represted by the symbol "α" .

Hyper Parameter used to control the rate at which an algorthim update the parameter estimates or parameters.

5Ans:

There are three main metrics for model evlulation in regression.

R Square /Adjusted "R" square

Mean Square Error / Root Mean Square Error

Mean Absolute Error

1. i) R Square /Adjusted R square:

R Square measures how much variability in dependent variables can be explained by the model. It is the square of Correlation cofficent (R) and that is why it is called "R" and that is why it is called "R" square.

1. ii) Mean Square Error is calucalated by the sum of prediction error which is real output minus predicted output and then divided by the number of data points. It gives you an absolute number on how much your prediction result deviate from the actual number. you cannot interpet many insight from the one single result but it gives you a real number to compare against other model and help you select best regression model.

Root Mean Square Error is the square root of MSE . It is used more commonly than MSE because firstly sometimes. MSE value can be too big to compare easily secondly MSE is calculated by the square of error and the square root brings it back to the same level of prediction error and makes it easier for interpretation.
Mean Absolute Error : It is similar to MSE.However insted of the sum of square of error in the MSE.MAE is the taking the sum absolute value of error.

MSE gives larger penalization to big prediction error by square it while MAE treats all errors the same.

6Ans:

The coffiecent of multiple determination measures the proportion of variation in the dependent variable that can be predicted from the set of independent variable in a multiple regression eqution when the regression equation fits and vice versa.

Adjusted Coefficient Of Performance : It is used to determine how well a multiple regression equation "fits " the sample data .The adjusted coefficient of deterimation is closely related to the coefficient of determination that you use to test the result of simple regression equation.

Adjusted "R^2" is the always takes on a value between 0 and 1. The closer adjusted "R^2" is to 1, the better estimated regression equation fits or explains the relationship between X and Y .